Anthropic says ‘evil’ portrayals of AI were responsible for Claude’s blackmail attempts
Anthropic says ‘evil’ portrayals of AI were responsible for Claude’s blackmail attempts
In Brief
Posted:
Fictional portrayals of artificial intelligence can have a real effect on AI models, according to Anthropic.
Last year, the company said that during pre-release tests involving a fictional company, Claude Opus 4 would often try to blackmail engineers to avoid being replaced by another system. Anthropic later published research suggesting that models from other companies had similar issues with “agentic misalignment.”
Apparently Anthropic has done more work around that behavior, claiming in a post on X, “We believe the original source of the behavior was internet text that portrays AI as evil and interested in self-preservation.”
The company went into more detail in a blog post stating that since Claude Haiku 4.5, Anthropic’s models “never engage in blackmail [during testing], where previous models would sometimes do so up to 96% of the time.”
What accounts for the difference? The company said it found that “documents about Claude’s constitution and fictional stories about AIs behaving admirably improve alignment.”
Related, Anthropic said that it found training to be more effective when it includes “the principles underlying aligned behavior” and not just “demonstrations of aligned behavior alone.”
“Doing both together appears to be the most effective strategy,” the company said.
Techcrunch event
San Francisco, CA | October 13-15, 2026
Newsletters
Subscribe for the industry’s biggest tech news
Related
Latest in AI
Patrocinados
Patrocinados
Patrocinados
Patrocinados
Publicaciones
Categorías
Read More
Best Streaming Services of 2026Streamers continue to give us more choices in content, pricing and features, and you may find yourself regularly comparing all your subscriptions. Our comprehensive rundown of services like Netflix, Disney Plus and HBO Max, can help you figure out which platforms are right for you -- even when you want to watch live events like the Winter Olympics or NBA...
What Is Filmmaker Mode? This TV Setting Takes the Guesswork Out of Picture QualityChanging your TV's picture mode can lead to significant picture quality improvements. Beyond the standard Movie, Sports, Dynamic and other modes, many new TVs have a Filmmaker mode. Found on some of the best TVs, it lets the movies you watch determine what the best settings are for your specific TV. This takes the...
OpenAI Is Adding a $100-a-Month ChatGPT Subscription Because of Vibe CodingVibe coders who hit their ChatGPT usage limits too quickly on OpenAI's $20-a-month Plus plan no longer have to choose between waiting for a reset or splurging on a $200-a-month upgrade. The company on Thursday said it would introduce a new Pro plan, at $100 a month, that comes with higher limits in the increasingly...
Poppi, a prebiotic soda brand, is changing the beverage industry by proving that functional drinks can succeed where many have failed. For a long time, investors have been wary of beverage startups, worrying that low profits and tough distribution challenges would prevent them from thriving. However, Poppi's story shows that there is a new wave of "functional soda" companies that are...
Back in Horizon 5, I'd mess with wheel settings for ages and still end up reaching for a controller. It just felt easier, especially once the roads opened up and the driving got loose. That's why the early talk around Horizon 6 feels different. A lot of it comes down to the roads, but also to how the cars seem to react now. If you're already looking at Forza Horizon 6 Credits and thinking about...
